Export error with saved_model, pb, tflite

[ShubhamAtPhilips]

Search before asking

• I have searched the YOLOv5 issues and found no similar bug report.

YOLOv5 Component

Export

Bug

I trained a model using this models/hub/yolov3-tiny.yaml configuration. I am trying to convert it to TFlite.

I found out that I need to implement TFMaxPool2d and TFZeroPad2d in models/tf.py from this issue #7033.

I implemented it as follows in models/tf.py

class TFMaxPool2d(keras.layers.Layer):
    def __init__(self, k=(2, 2), s=None, p='same', w=None):
        super().__init__()
        self.m = keras.layers.MaxPool2D(pool_size=k, strides=s, padding='SAME')

    def call(self, inputs):
        return self.m(inputs)

class TFZeroPad2d(keras.layers.Layer):
    def __init__(self, p=(1, 1), w=None):
        super().__init__()
        print(p)
        self.z = keras.layers.ZeroPadding2D(padding=p)

    def call(self, inputs):
        return self.z(inputs)

Now, I am getting the error

TensorFlow SavedModel: starting export with tensorflow 2.7.1...
                 from  n    params  module                                  arguments                        0                -1  1       464  models.common.Conv                      [3, 16, 3, 1]                    1                -1  1         0  torch.nn.modules.pooling.MaxPool2d      [2, 2, 0]                        2                -1  1      4672  models.common.Conv                      [16, 32, 3, 1]                   3                -1  1         0  torch.nn.modules.pooling.MaxPool2d      [2, 2, 0]                        4                -1  1     18560  models.common.Conv                      [32, 64, 3, 1]                   5                -1  1         0  torch.nn.modules.pooling.MaxPool2d      [2, 2, 0]                        6                -1  1     73984  models.common.Conv                      [64, 128, 3, 1]                  7                -1  1         0  torch.nn.modules.pooling.MaxPool2d      [2, 2, 0]                        8                -1  1    295424  models.common.Conv                      [128, 256, 3, 1]                 9                -1  1         0  torch.nn.modules.pooling.MaxPool2d      [2, 2, 0]                       10                -1  1   1180672  models.common.Conv                      [256, 512, 3, 1]               

TensorFlow SavedModel: export failure: `padding` should have two elements. Received: [0, 1, 
0, 1].

TensorFlow Lite: starting export with tensorflow 2.7.1...

TensorFlow Lite: export failure: 'NoneType' object has no attribute 'call'

Can someone help me implement 4-tuple padding TF/Keras equivalent of torch.nn.ZeroPad2d?

Thank you.

Environment

No response

Minimal Reproducible Example

No response

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[github-actions]

👋 Hello @ShubhamAtPhilips, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://ultralytics.com or email support@ultralytics.com.

Requirements

Python>=3.7.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab and Kaggle notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training (train.py), validation (val.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu every 24 hours and on every commit.

[ShubhamAtPhilips]

Update,

I updated the function definitions, and it seems to work now.

class TFMaxPool2d(keras.layers.Layer):
    def __init__(self, k=(2, 2), s=None, p='same', w=None):
        super().__init__()
        if p == 0:
            padding = 'valid'
        else:
            padding = 'same'
        self.m = keras.layers.MaxPool2D(pool_size=k, strides=s, padding=padding)
        
    def call(self, inputs):
        return self.m(inputs)

class TFZeroPad2d(keras.layers.Layer):
    def __init__(self, p=(1, 1), w=None):
        super().__init__()
        if hasattr(p, '__len__'):
            if len(p) == 4:
                p = ((p[2], p[3]), (p[0], p[1]))
        self.z = keras.layers.ZeroPadding2D(padding=p)

    def call(self, inputs):
        return self.z(inputs)

Can someone verify if it's correct? If yes, I can submit a PR.

[glenn-jocher]

@ShubhamAtPhilips oh cool! Yes go ahead and submit a PR. To verify your implementation export yolov3-tiny.pt to TFLite or SavedModel format and then run val.py on your exported model and compare against PyTorch model. Both should have identical mAP to several decimal places.

[github-actions]

👋 Hello, this issue has been automatically marked as stale because it has not had recent activity. Please note it will be closed if no further activity occurs.

Access additional YOLOv5 🚀 resources:

• Wiki – https://github.com/ultralytics/yolov5/wiki
• Tutorials – https://docs.ultralytics.com/yolov5
• Docs – https://docs.ultralytics.com

Access additional Ultralytics ⚡ resources:

• Ultralytics HUB – https://ultralytics.com/hub
• Vision API – https://ultralytics.com/yolov5
• About Us – https://ultralytics.com/about
• Join Our Team – https://ultralytics.com/work
• Contact Us – https://ultralytics.com/contact

Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

Thank you for your contributions to YOLOv5 🚀 and Vision AI ⭐!

[wakasu]

This works, i was able to convert yolov3-tiny to tflite 👍

[glenn-jocher]

@wakasu that's great to hear! 🎉 If you have any further questions or need assistance with anything else, feel free to ask. Good luck with your project!